# Changelog

## 2.1.0 - 2026-08-09

- Make model catalog fetching resilient: per-attempt timeouts, retries with backoff for transient failures (timeouts, network errors, HTTP 429/5xx, invalid responses), and fast failure on permanent errors such as a bad API key.
- Refresh the catalog through pi's `refreshModels` hook at startup and after `/login` instead of only at extension load.
- Persist the last-known-good catalog and restore it when the API is unreachable, keeping models selectable during outages.
- Always register the provider so it stays visible in `/login` even when the API is down.
- `/cortecs-models` now lists the live catalog instead of the load-time snapshot.
- Bump the `@earendil-works/pi-coding-agent` devDependency from `^0.79.0` to `^0.84.0` (required for the new `refreshModels` API).

## 2.0.1 - 2026-06-22

- Ship the updated README (drop the redundant pi-install section, add an author section).
- Published through CI with a provenance attestation.

## 2.0.0 - 2026-06-22

- **Renamed the package to `@czottmann/pi-cortecs`.** Install with `pi install npm:@czottmann/pi-cortecs`. The old unscoped `pi-cortecs` package is deprecated.
- Ship the extension as raw TypeScript (`extensions/cortecs.ts`) loaded directly by pi, instead of a compiled `dist/` build. No build step is needed for local installs.

## 1.1.0 - 2026-06-14

- Update the Cortecs provider API key reference to pi's current `$CORTECS_API_KEY` syntax, removing the startup deprecation warning.

## 1.0.0 - 2026-05-08

Initial release.

- Registers Cortecs as a pi provider.
- Fetches the Cortecs model catalog on startup.
- Registers tool-capable models with context, pricing, image, and reasoning metadata.
- Supports pi's API-key login flow and the `CORTECS_API_KEY` environment variable.
- Adds `/cortecs-models` to list available Cortecs models.